Lightsim2grid

--------------------
- [BREAKING] the behaviour of the `newton_pf` function is not
consistent with pandapower default concerning distributed slack.
- [FIXED] an issue in the distributed slack case spotted by pandapower team
thanks to them (see https://github.com/e2nIEE/pandapower/pull/1455)
- [IMPROVED] lightsim2grid will now use the single slack algorithm if the
grids counts only one slack bus (performance increase)

-------------------
- [BREAKING] change the interface of the `newton_pf` function to reflect pandapower change in their
latest version (arguments `ref` has been added). You can still use the old `newton_pf` function, with the
old signature by importing `newtonpf_old` instead or explicitly importing the new one by importing `newtonpf_new`
- [BREAKING] `SecurityAnalysis` now also returns the active flows when calling `security_analysis.get_flows()`
- [BREAKING] change the file names (python side) to be compliant with pep 8. You can no longer
do things like `from lightsim2grid.LightSimBackend import LightSimBackend` change it to
`from lightsim2grid import LightSimBackend` (preferred method)
- [BREAKING] change the file names (python side) to be compliant with pep 8. You can no longer
do things like `from lightsim2grid.initGridModel import init` change it to
`from lightsim2grid.gridmodel import init` (preferred method) (same for `GridModel` class)
- [FIXED] a bug that lead to the wrong computation of the dc powerflow in case of `sn_mva != 1.` and phase shifters.
- [FIXED] bug preventing to use the NICSLU linear solver in the `GridModel`
- [FIXED] compilation warnings on clang (missing virtual destructor, unused variables, etc.)
- [FIXED] a bug in the `SecurityAnalysisCPP`: when it diverges for some contingencies, the others were not simulated properly.
- [FIXED] `LightSimBackend` now contains members for `shunts` and `***_theta` as it does for the other quantities. This improves the consistency, but most importantly
fixes some bugs when used in earlier grid2op versions
- [ADDED] possibility to compute the active flows using the `BaseMultiplePower`
- [ADDED] possibility to change linear solver used when performing a DC solver
- [ADDED] possibility to make powerflow with distributed slack bus (only for newton raphson at the moment)
- [ADDED] access (read only) to the element of a lightsim2grid grid with the `get_XXX` (*eg* `get_loads()`) methods (see documentation)
- [ADDED] direct access to the solver used in the grid model python side
- [ADDED] unittest in circleci.
- [ADDED] all kind of solvers based on different linear solvers (Eigen sparse LU, KLU or NICSLU) for Newton Raphson and
DC approximation (9 solvers in total)
- [IMPROVED] use of `steady_clock` to retrieve the ellapse time c++ side
- [IMPROVED] refactoring of the c++ part to use template mecanism instead of inheritance for the
Newton Raphson and DC solvers.
- [IMPROVED] `GridModel` now contains two different solvers, one for AC powerflow and one for DC powerflow.
- [IMPROVED] error message in the solver are now embedded in an Enum instead of being integers, for better readibility.
- [IMPROVED] error message when the powerflow diverge (error are read from c++ now)

-------------------
- [ADDED] possibility to perform dc powerflow
- [ADDED] a class to compute flows on whole time series when the Ybus does not change (see `TimeSerie`)
- [ADDED] a class to compute flows on multiple contingencies, when Sbus does not change (see `SecurityAnalysis`).
- [IMPROVED] running speed of Newton Raphson solvers with better filling of sparse matrices
- [IMPROVED] upgrade to SuiteSparse `v5.10.1`
- [IMPROVED] upgrade to eigen `3.4.0` (stable release)
- [IMPROVED] clean the compilation warnings on microsoft windows (force the conversion from
`Eigen::EigenBase<Derived>::Index` to `int` using `static_cast`)
- [IMPROVED] add the proper optimization flag for windows (`/O2` instead of `-03` on linux / macos)
- [IMPROVED] high performance gain when topology is not changed between steps (gain obtained by
reusing the previous Ybus)

------------------
- [FIXED] a bug for static generator (wrong signed convention were used in some part of the c++ code). This has
no impact at all for provided grid2op environments.
- [FIXED] An issue where the backend could get "stuck" in a wrong state because of the way the Vinit was computed (see
`Issue 30 <https://github.com/BDonnot/lightsim2grid/issues/30>`_)
- [ADDED] experimental support for the `NICSLU` linear solver (requires a proper license and library, see
https://github.com/chenxm1986/nicslu for more information. Support does not include multi threaded at the moment).
- [IMPROVED] minor performance improvements for the solvers based on Newton Raphson (faster filling of the Jacobian
matrix after the first iteration)

-------------------
- [FIXED] minor issues in the benchmark (some time measurments were wrong)
- [ADDED] lightsim2grid package now can be distributed on pypi
- [ADDED] compilation of SuiteSparse using cmake
- [ADDED] compatibility with the KLU linear solver on windows based systems.
- [IMPROVED] the package should now be available on pypi

-------------------
- [FIXED] `GridModel` now properly throw "out_of_range" exception when trying to change the bus of non existing
elements
- [FIXED] wrong units were displayed for the iterators for lines and transformers.
- [ADDED] now able to retrieve the powerlines parameters python side.
- [IMPROVED] more explicit error messages when the building of the `Ybus` matrix fails.
- [IMPROVED] now the solver is not reset when using the `backend._grid.check_solution`
- [IMPROVED] upgrade SuiteSparse to version `v5.10.1`
- [IMPROVED] upgrade eigen to version `3.4-rc1`